Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

IEEE International Conference on Enabling Technologies: Infrastructure for Collaborative Enterprises (WETICE)

Stochastic Quantum Neural Networks for Collaborative Anomaly Detection in Distributed Enterprise Systems

Gautier-Edouard Filardo

Published Jul 1, 2026• 0 citations
Publisher

Abstract

Collaborative enterprise systems generate massive data streams requiring real-time anomaly detection. This paper extends our Stochastic Quantum Neural Network (QNNS) formalism to collaborative anomaly detection, leveraging superposition for parallel hypothesis testing, entanglement for inter-node correlation modeling, and stochastic noise for robustness. We validate three theoretical predictions through four experiments across two circuit scales (8 and 12 qubits), then quantify the theory–practice gap through bound tightness analysis, scaling diagnostics, and entanglement advantage decomposition. Our central result demonstrates that ring-topology entanglement is essential for detecting non-local correlations: on an XOR-structured anomaly dataset, ring entanglement achieves 84.5% accuracy versus 49.9% for circuits without entanglement (p < 0.01), with the ring closure alone accounting for +6.0% accuracy gain. The adversarial robustness bound |∆A| ≤ ϵ/(γ + ϵ) is satisfied at all test points, but tightness analysis reveals it overestimates vulnerability by 3–100×, characterizing it as a conservative safety guarantee rather than a tight prediction. A scaling study across 8-qubit (24 parameters) and 12-qubit (60 parameters) circuits reveals that the noise-as-regularization effect does not emerge at either scale; critically, the mean regularization deficit increases from +0.018 (8q) to +0.032 (12q) with a positive scaling slope, confirming a fundamental mechanism failure rather than a scale limitation. Together, these experiments and analyses transform the QNNS framework from a theoretical proposal to a partially validated architecture with precisely quantified strengths and honestly characterized limitations.

DOI
10.1109/wetice70595.2026.11772195

Related papers

A Distributed Quantum Neural Network for Network Anomaly Detection

Moe Hdaib, Sutharshan Rajasegarar, Seng W. Loke

The scalability of quantum neural networks (QNNs) is limited by hardware constraints such as restricted qubit connectivity and circuit depth. This paper presents a distributed quantum neural network (DQNN) architecture that enables multi-processor quantum computation through teleportationbased distributed CNOT gates. The proposed framework allows variational circuits to operate seamlessly across nodes using ancilla-mediated entanglement and pre-shared Bell pairs, without requiring direct physical links. We demonstrate that this distributed design inherently mitigates the barren plateau problem by reducing circuit depth per node, localizing cost functions, and introducing structured inter-node entanglement, thereby improving gradient trainability. Experiments on network traffic anomaly detection show that the DQNN achieves competitive classification accuracy while efficiently managing resource tradeoffs, including a 33 % ancilla overhead, minimal communication cost (2 bits per distributed CNOT), and distribution of up to 50 % of CNOT operations. These results establish a scalable and trainable architecture for near-term quantum machine learning.

International Conference on Quantum Communications, Networking, and Computing (QCNC)•Apr 6, 2026•0 citations

Decoherence as Defence and the Magnitude of Noise Regularisation: A Rigorous N -Qubit Theory of Stochastic Quantum Neural Networks for Adversarially Robust Network Intrusion Detection

Gautier-Edouard Filardo

Stochastic quantum neural networks (SQNNs) encode neuronal activations as qubits, synaptic topology as entanglement, and neural noise through a Lindblad master equation. A recent conference study applied a ring-entangled SQNN to collaborative intrusion detection and reached three conclusions: ring entanglement is \emph{essential} for non-local anomaly detection; an adversarial-resilience bound holds but is \emph{conservative}; and the depolarising channel \emph{fails} to act as a dropout-style regulariser, behaving instead as output noise. It left open whether a per-gate stochastic deactivation (``true quantum dropout'') could regularise where the depolarising channel could not, and whether the loose robustness bound could be replaced by a predictive theory. This paper resolves both and extends the framework to real data and to neutral-atom hardware. We give an $N$-qubit formulation through the stochastic master equation and its vectorised Liouvillian, and prove a \emph{decoherence-contraction theorem}: a depolarising channel of strength $γ$ over $L$ entangling layers contracts every weight-$w$ Pauli read-out by a factor $(1-4γ/3)^{wL}$ (for the weight-$1$ read-out used here, $(1-4γ/3)^{L}$); building on the general noise-as-defence result of Du et al., we make this quantitative and operational for intrusion detection. On the real NSL-KDD dataset under white-box FGSM and PGD attacks, a depolarising SQNN trained with the channel is, over seven seeds under strong $\ell_\infty$/$\ell_2$ attacks, significantly more robust than the noiseless circuit ($\ell_\infty$ PGD-$20$, $p=0.04$, large effect) and, critically, never suffers the catastrophic robustness collapse that the noiseless model and gradient-trained classical detectors (which fall from $95\%$ to $47\%$) do, cutting robustness variance roughly twofold; we show this robustness arises from a noise-reshaped training boundary rather than from attack-time gradient contraction. For generalisation, we derive an adaptive-penalty formula showing that per-gate dropout implements a curvature-weighted $L_2$ penalty $\tfrac{p(1-p)}{2}\sumθ^2\partial^2_θL$ in weight space, maximised at $p=1/2$, whereas depolarising noise implements an output-space penalty. A $30$-seed study confirms the formula's quantitative prediction: both mechanisms reduce the train-test gap by a small but statistically significant margin ($\approx\!0.01$; $p<10^{-4}$ and $p=0.004$), are statistically indistinguishable from each other, and the effect is concentrated where overfitting is largest; increasing the dropout rate past $1/2$ does not help, as the formula predicts. The single-seed dichotomy of prior work does not survive replication. We close with a neutral-atom realisation and a feasibility-by-$N$ analysis.

arXiv:2606.24219•Jun 23, 2026•1 citation

Quantum Machine Learning for Anomaly Detection: The Future of Smarter and Safer IoT Networks

Ishtiaq Ahmad, Liang Yang, Yazeed Masaud Alkhrijah +4 more

The proliferation of large-scale Internet of Things (IoT) networks introduce significant challenges for real-time anomaly detection, particularly due to the massive volume, high speed, and heterogeneity of IoT-generated data. Traditional and deep learning-based models, while effective to some extent, often face computational bottlenecks that limit their scalability and responsiveness in complex IoT environments. To address these limitations, this paper investigates the application of Quantum Machine Learning (QML) for anomaly detection in IoT systems. By leveraging the parallelism and high-dimensional processing capabilities of quantum computing, QML offers a promising pathway to overcome current computational constraints. We propose a novel anomaly detection framework based on Quantum Generative Adversarial Networks (QGANs) designed to enhance security, reliability, and real-time responsiveness across diverse IoT scenarios, including smart homes, industrial IoT, healthcare monitoring, and critical infrastructure protection. Furthermore, this paper provides a comprehensive analysis of existing anomaly detection challenges, practical IoT applications, and outlines future research directions, including hybrid quantum-classical architectures, resource-efficient quantum algorithms for edge computing, and the integration of QML with emerging technologies to ensure scalable, secure, and robust IoT systems.

IEEE Network•Jan 12, 2026•4 citations

Adaptive Quantum Entanglement Networks for Real-Time Financial Fraud Detection: A Novel Framework with Temporal Correlation Analysis

Yalla Jnan Devi Satya Prasad

Financial fraud detection faces unprecedented challenges due to increasing transaction volumes, sophisticated attack vectors, and the need for real-time processing. This paper introduces Adaptive Quantum Entanglement Networks (AQEN), a revolutionary quantum machine learning framework that leverages temporal correlations and dynamic entanglement for superior fraud detection performance. Under Assumptions 1-3 (Section 2), our theoretical analysis predicts asymptotic improvements that can scale exponentially in qubit count compared to naive classical representations for structured temporal models; empirical results validate these gains in realistic simulation regimes We establish rigorous theoretical foundations through novel mathematical frameworks, proving quantum advantage bounds for temporal fraud detection and deriving explicit performance guarantees. The core innovation lies in adaptive entanglement mechanisms that dynamically adjust quantum correlations based on transaction patterns, creating quantum representations that mirror real-world fraud network structures. Experimental validation on comprehensive financial datasets demonstrates that AQEN achieves 94.7% accuracy, 81.2% precision, and 79.8% recall, representing significant improvements over classical state-of-the-art methods (93.5% accuracy, 78.5% precision, 74.2% recall for XGBoost). Statistical analysis confirms the improvements are highly significant (p < 0.001), with AQEN achieving AUC = 0.951 compared to 0.923 for the best classical approach. The paper provides complete algorithmic specifications, quantum circuit implementations, and comprehensive performance analysis, establishing AQEN as a breakthrough approach for next-generation fraud detection systems with practical deployment potential on nearterm quantum devices.

•Sep 10, 2025•0 citations

Quantum-enhanced spiking intelligence framework for real-time anomaly detection in industrial internet of things

Kamran Ahmad Awan, Ashwag Albakri, Fatimah Alhayan +4 more

Industrial Internet of Things (IIoT) anomaly detection imposes concurrent requirements for temporal consistency, computational latency control, and representation of high-dimensional heterogeneous sensor data. Addressing these constraints requires models capable of processing non-stationary streams while maintaining bounded inference time and preserving inter-sensor dependencies. The proposed Quantum-Enhanced Spiking Neural Network (QESNN) integrates parameterized quantum circuits with event-driven spiking computation within a unified inference structure. The architecture is comprised of five computation modules. The Quantum Neurons for Sensor Data Fusion module maps multimodal sensor observations onto amplitude-phase quantum states and then uses nonlinear unitary transformations to generate joint feature representations through multiple channels. The Entanglement-Based Device Synchronization module aligns phases between nodes by using entangled states, coherence-control operators, and correlations for synchronization. The Probabilistic Quantum Spiking for Anomaly Detection module uses spike probability based on tunneling and incorporates fidelity and entropy metrics to estimate the likelihood of anomalies under uncertainty. The Adaptive Quantum Learning for Dynamic Optimization module tunes model parameters via time-dependent Hamiltonian dynamics with annealing strategies and feedback correction schemes to ensure robust convergence in dynamic environments. Finally, the Quantum Decoherence Management for Reliable Processing module models the impact of noise, utilizes coherence-control operations, and conducts reliability checks in order to maintain stable quantum-state computation amidst external disturbances. When tested on the MVTec Anomaly Detection dataset, the model achieves classification accuracy of 98.7% with 512 logical quantum-spiking units and an inference time of 3.29 s. The effect of each component is estimated via ablation study - eliminating the Quantum Neurons for Sensor Data Fusion module leads to classification accuracy of 92.5%, whereas elimination of the Quantum Decoherence Management for Reliable Processing module produces the lowest classification accuracy of 88.9%.

Scientific Reports•May 30, 2026•1 citation

A Comprehensive Analysis of Accuracy and Robustness in Quantum Neural Networks

Ban Q. Tran, Duong M. Chu, Hai T. D. Pham +3 more

Quantum Machine Learning (QML) has recently emerged as a highly promising research frontier. Within this domain, Quantum Neural Networks (QNNs),characterized by Variational Quantum Circuits (VQCs) at their core and featuring layers of quantum gates optimized by classical algorithms, have garnered significant attention. However, a rigorous and exhaustive evaluation of their practical performance remains largely incomplete. In this study, we conduct a comprehensive comparative analysis of three prominent hybrid classical-quantum architectures: Quantum Convolutional Neural Networks (QCNN), Quantum Recurrent Neural Networks (QRNN), and Quantum Vision Transformers (QViT), focusing on the critical dimensions of generalization, accuracy, and robustness. Our findings provide novel insights that address previous evaluative gaps. Notably, while these models exhibit exceptional performance on low-feature datasets such as MNIST, their learning efficacy degrades significantly when transitioned to high-feature datasets. Furthermore, convolutional-based models like QCNN appear less effective on high-dimensional data than other machine learning architectures. Additionally, while all models are susceptible to adversarial noise, traditional architectures, such as recurrent and convolutional networks, demonstrate superior resilience. Conversely, in the presence of quantum noise, the transformer-based architecture proves its strength by maintaining high robustness against measurement noise, channel noise, and finite-shot effects, whereas other architectures suffer marked performance declines. These results provide a granular perspective on the current state of the field and underscore the critical importance of tailoring model selection to the constraints of contemporary Noisy Intermediate-Scale Quantum (NISQ) environments.

quant-pharXiv:2604.26110•Apr 30, 2026•1 citation
Collaborative enterprise systems generate massive data streams requiring real-time anomaly detection. This paper extends our Stochastic Quantum Neural Network (QNNS) formalism to collaborative anomaly detection, leveraging superposition for parallel hypothesis testing, entanglement for inter-node correlation modeling, and stochastic noise for robustness. We validate three theoretical predictions through four experiments across two circuit scales (8 and 12 qubits), then quantify the theory–practice gap through bound tightness analysis, scaling diagnostics, and entanglement advantage decomposition. Our central result demonstrates that ring-topology entanglement is essential for detecting non-local correlations: on an XOR-structured anomaly dataset, ring entanglement achieves 84.5% accuracy versus 49.9% for circuits without entanglement (p < 0.01), with the ring closure alone accounting for +6.0% accuracy gain. The adversarial robustness bound |∆A| ≤ ϵ/(γ + ϵ) is satisfied at all test points, but tightness analysis reveals it overestimates vulnerability by 3–100×, characterizing it as a conservative safety guarantee rather than a tight prediction. A scaling study across 8-qubit (24 parameters) and 12-qubit (60 parameters) circuits reveals that the noise-as-regularization effect does not emerge at either scale; critically, the mean regularization deficit increases from +0.018 (8q) to +0.032 (12q) with a positive scaling slope, confirming a fundamental mechanism failure rather than a scale limitation. Together, these experiments and analyses transform the QNNS framework from a theoretical proposal to a partially validated architecture with precisely quantified strengths and honestly characterized limitations.
The scalability of quantum neural networks (QNNs) is limited by hardware constraints such as restricted qubit connectivity and circuit depth. This paper presents a distributed quantum neural network (DQNN) architecture that enables multi-processor quantum computation through teleportationbased distributed CNOT gates. The proposed framework allows variational circuits to operate seamlessly across nodes using ancilla-mediated entanglement and pre-shared Bell pairs, without requiring direct physical links. We demonstrate that this distributed design inherently mitigates the barren plateau problem by reducing circuit depth per node, localizing cost functions, and introducing structured inter-node entanglement, thereby improving gradient trainability. Experiments on network traffic anomaly detection show that the DQNN achieves competitive classification accuracy while efficiently managing resource tradeoffs, including a 33 % ancilla overhead, minimal communication cost (2 bits per distributed CNOT), and distribution of up to 50 % of CNOT operations. These results establish a scalable and trainable architecture for near-term quantum machine learning.
Stochastic quantum neural networks (SQNNs) encode neuronal activations as qubits, synaptic topology as entanglement, and neural noise through a Lindblad master equation. A recent conference study applied a ring-entangled SQNN to collaborative intrusion detection and reached three conclusions: ring entanglement is \emph{essential} for non-local anomaly detection; an adversarial-resilience bound holds but is \emph{conservative}; and the depolarising channel \emph{fails} to act as a dropout-style regulariser, behaving instead as output noise. It left open whether a per-gate stochastic deactivation (``true quantum dropout'') could regularise where the depolarising channel could not, and whether the loose robustness bound could be replaced by a predictive theory. This paper resolves both and extends the framework to real data and to neutral-atom hardware. We give an NNN-qubit formulation through the stochastic master equation and its vectorised Liouvillian, and prove a \emph{decoherence-contraction theorem}: a depolarising channel of strength γγγ over LLL entangling layers contracts every weight-www Pauli read-out by a factor (1−4γ/3)wL(1-4γ/3)^{wL}(1−4γ/3)wL (for the weight-111 read-out used here, (1−4γ/3)L(1-4γ/3)^{L}(1−4γ/3)L); building on the general noise-as-defence result of Du et al., we make this quantitative and operational for intrusion detection. On the real NSL-KDD dataset under white-box FGSM and PGD attacks, a depolarising SQNN trained with the channel is, over seven seeds under strong ℓ∞\ell_\inftyℓ∞​/ℓ2\ell_2ℓ2​ attacks, significantly more robust than the noiseless circuit (ℓ∞\ell_\inftyℓ∞​ PGD-202020, p=0.04p=0.04p=0.04, large effect) and, critically, never suffers the catastrophic robustness collapse that the noiseless model and gradient-trained classical detectors (which fall from 95%95\%95% to 47%47\%47%) do, cutting robustness variance roughly twofold; we show this robustness arises from a noise-reshaped training boundary rather than from attack-time gradient contraction. For generalisation, we derive an adaptive-penalty formula showing that per-gate dropout implements a curvature-weighted L2L_2L2​ penalty p(1−p)2∑θ2∂θ2L\tfrac{p(1-p)}{2}\sumθ^2\partial^2_θL2p(1−p)​∑θ2∂θ2​L in weight space, maximised at p=1/2p=1/2p=1/2, whereas depolarising noise implements an output-space penalty. A 303030-seed study confirms the formula's quantitative prediction: both mechanisms reduce the train-test gap by a small but statistically significant margin (≈ ⁣0.01\approx\!0.01≈0.01; p<10−4p<10^{-4}p<10−4 and p=0.004p=0.004p=0.004), are statistically indistinguishable from each other, and the effect is concentrated where overfitting is largest; increasing the dropout rate past 1/21/21/2 does not help, as the formula predicts. The single-seed dichotomy of prior work does not survive replication. We close with a neutral-atom realisation and a feasibility-by-NNN analysis.
The proliferation of large-scale Internet of Things (IoT) networks introduce significant challenges for real-time anomaly detection, particularly due to the massive volume, high speed, and heterogeneity of IoT-generated data. Traditional and deep learning-based models, while effective to some extent, often face computational bottlenecks that limit their scalability and responsiveness in complex IoT environments. To address these limitations, this paper investigates the application of Quantum Machine Learning (QML) for anomaly detection in IoT systems. By leveraging the parallelism and high-dimensional processing capabilities of quantum computing, QML offers a promising pathway to overcome current computational constraints. We propose a novel anomaly detection framework based on Quantum Generative Adversarial Networks (QGANs) designed to enhance security, reliability, and real-time responsiveness across diverse IoT scenarios, including smart homes, industrial IoT, healthcare monitoring, and critical infrastructure protection. Furthermore, this paper provides a comprehensive analysis of existing anomaly detection challenges, practical IoT applications, and outlines future research directions, including hybrid quantum-classical architectures, resource-efficient quantum algorithms for edge computing, and the integration of QML with emerging technologies to ensure scalable, secure, and robust IoT systems.
Financial fraud detection faces unprecedented challenges due to increasing transaction volumes, sophisticated attack vectors, and the need for real-time processing. This paper introduces Adaptive Quantum Entanglement Networks (AQEN), a revolutionary quantum machine learning framework that leverages temporal correlations and dynamic entanglement for superior fraud detection performance. Under Assumptions 1-3 (Section 2), our theoretical analysis predicts asymptotic improvements that can scale exponentially in qubit count compared to naive classical representations for structured temporal models; empirical results validate these gains in realistic simulation regimes We establish rigorous theoretical foundations through novel mathematical frameworks, proving quantum advantage bounds for temporal fraud detection and deriving explicit performance guarantees. The core innovation lies in adaptive entanglement mechanisms that dynamically adjust quantum correlations based on transaction patterns, creating quantum representations that mirror real-world fraud network structures. Experimental validation on comprehensive financial datasets demonstrates that AQEN achieves 94.7% accuracy, 81.2% precision, and 79.8% recall, representing significant improvements over classical state-of-the-art methods (93.5% accuracy, 78.5% precision, 74.2% recall for XGBoost). Statistical analysis confirms the improvements are highly significant (p < 0.001), with AQEN achieving AUC = 0.951 compared to 0.923 for the best classical approach. The paper provides complete algorithmic specifications, quantum circuit implementations, and comprehensive performance analysis, establishing AQEN as a breakthrough approach for next-generation fraud detection systems with practical deployment potential on nearterm quantum devices.
Industrial Internet of Things (IIoT) anomaly detection imposes concurrent requirements for temporal consistency, computational latency control, and representation of high-dimensional heterogeneous sensor data. Addressing these constraints requires models capable of processing non-stationary streams while maintaining bounded inference time and preserving inter-sensor dependencies. The proposed Quantum-Enhanced Spiking Neural Network (QESNN) integrates parameterized quantum circuits with event-driven spiking computation within a unified inference structure. The architecture is comprised of five computation modules. The Quantum Neurons for Sensor Data Fusion module maps multimodal sensor observations onto amplitude-phase quantum states and then uses nonlinear unitary transformations to generate joint feature representations through multiple channels. The Entanglement-Based Device Synchronization module aligns phases between nodes by using entangled states, coherence-control operators, and correlations for synchronization. The Probabilistic Quantum Spiking for Anomaly Detection module uses spike probability based on tunneling and incorporates fidelity and entropy metrics to estimate the likelihood of anomalies under uncertainty. The Adaptive Quantum Learning for Dynamic Optimization module tunes model parameters via time-dependent Hamiltonian dynamics with annealing strategies and feedback correction schemes to ensure robust convergence in dynamic environments. Finally, the Quantum Decoherence Management for Reliable Processing module models the impact of noise, utilizes coherence-control operations, and conducts reliability checks in order to maintain stable quantum-state computation amidst external disturbances. When tested on the MVTec Anomaly Detection dataset, the model achieves classification accuracy of 98.7% with 512 logical quantum-spiking units and an inference time of 3.29 s. The effect of each component is estimated via ablation study - eliminating the Quantum Neurons for Sensor Data Fusion module leads to classification accuracy of 92.5%, whereas elimination of the Quantum Decoherence Management for Reliable Processing module produces the lowest classification accuracy of 88.9%.
Quantum Machine Learning (QML) has recently emerged as a highly promising research frontier. Within this domain, Quantum Neural Networks (QNNs),characterized by Variational Quantum Circuits (VQCs) at their core and featuring layers of quantum gates optimized by classical algorithms, have garnered significant attention. However, a rigorous and exhaustive evaluation of their practical performance remains largely incomplete. In this study, we conduct a comprehensive comparative analysis of three prominent hybrid classical-quantum architectures: Quantum Convolutional Neural Networks (QCNN), Quantum Recurrent Neural Networks (QRNN), and Quantum Vision Transformers (QViT), focusing on the critical dimensions of generalization, accuracy, and robustness. Our findings provide novel insights that address previous evaluative gaps. Notably, while these models exhibit exceptional performance on low-feature datasets such as MNIST, their learning efficacy degrades significantly when transitioned to high-feature datasets. Furthermore, convolutional-based models like QCNN appear less effective on high-dimensional data than other machine learning architectures. Additionally, while all models are susceptible to adversarial noise, traditional architectures, such as recurrent and convolutional networks, demonstrate superior resilience. Conversely, in the presence of quantum noise, the transformer-based architecture proves its strength by maintaining high robustness against measurement noise, channel noise, and finite-shot effects, whereas other architectures suffer marked performance declines. These results provide a granular perspective on the current state of the field and underscore the critical importance of tailoring model selection to the constraints of contemporary Noisy Intermediate-Scale Quantum (NISQ) environments.